Hi,

the last month i have been trying to get a firmer grip on splitting. I have studied positions from 3-ply money games (without Jacoby rule) against gnu. I was able to formulate rules which apply to almost all positions.
However i came across 3 positions that seem to contradict this rules:

Here are some rules:

1. Against a better board only split against one hitting number.
2. Don't build and split (this is only good when the split is safer than normal).
3. It is possible to safety the split by inactivating a hitter, i call this pressure splitting.

In position 697 o (white) has a better board so splitting 24/20 is riskier than normal, thats why my play would be 24/20 and 24/23 thus inactivating the hitter on the 17-point. Gnu however suggests 24/23 13/9 (equity difference 0,033).

In position 1073 I would play 13/8 since the split is already riskier than normal, on later sight I even thought according to my rules that 24/23 and 13/9 might be better. Gnu however suggests 13/9 24/22 (0,003 better than 13/8).

In my "splitting theory" I also included the coming under the gun of a single back man, I saw that most of the time the risk is comparable with the risk of splitting, though one gradation riskier.

In position 746 the "pseudo split" is already riskier than normal since the white board is better, and its a single checker. So I would play 13/8. Gnu suggests 13/10 24/22 (difference= 0,008).

These 3 positions are exceptions to my rules (lets say 3 odds out of 100 normals) and my question to the forum members is why??

The equity differences, especially for 1073 and 746 are pretty small so a rollout might show change the evals slightly.

Regarding 1073 I don't see a lot of downside to 24/22 as it gives the same likelihood of being pointed on, albeit making the 3 is probably a little better than making the 2 for o, on the other hand making the 22 anchor is better for x than the 23 if you anchor later.

I guess the real question is the risk of the blot on the 9 point. There are only 5 shots that hit that blot, and you have now added an extra builder for the higher inner board points, my thought being that you would break the 8 to make the 5 or 4. I suppose that could be debated, but if you did make an inner board point by breaking the eight you are looking at a reasonable number of returns if you get hit there, a good shot at an nice advanced anchor, and you are now working from a even board strength positions.

The 13/9 22/24 play seems to me to give you a great deal of flexibility in overall game plan. If you don't get hit on the 9 you have a reasonable opportunity to start making home board points and switching to attack mode. If you do get hit on the 9 you have a good shot at an advanced anchor for a holding game, plus the third checker back might well come in handy for timing and/or hitting, and again if you don't get hit on the 9 you still have a possible to convert into a running game with a couple of big rolls. I am not sure your other plays leave that much flexibility.

In 697 I think you are over generalizing the better board rule. 31 is the only roll that points on the 20 blot and other hits are going to leave substantial blotting numbers. Hanging back on the 1 allows o to bring builders down on the next roll with little risk so you could quickly find yourself sitting on the 24 point with a prime growing in front of you. With the checker on the 20 putting blots in the outfield for builders becomes very risky for o and at the same time you are threatening to grab control of the 20 point.

A couple of comments:

Of your 3 positions, only #1 is a real splitting problem. X is already split in #2, and X has only one man back in #3.

In #3, the issue is "Does X want to advance his back man in O's board or not?" The answer is usually "no", but there are plenty of exceptions.

In general, splitting is heavily dependent on the whole position, not just O's home board. You have to look carefully at what's happening on X's side of the board before you decide.

Sorry is this is a bit vague, but splitting is a very complicated topic.

You seem to be over valuing rule number one in the early game. You probably don't want to get into a blot hiting contest when the opponent has the better board, but it's not the same with splitting. In all of the positions the opponent is a major underdog to point on your head and lacks builders.

The number of builders/attackers is what you're not considering.

Suppose that the opponent opened up with 43 two checkers down. You get a 41 and now you slot, not just because of duplication and the fact that a conter prime is a good plan, but because is more dangerous to split even though he still only has a 1 pt board.

I think it's pretty easy to know when to split in the early game if you play enough. Basically with every non constructive roll or any non good hitting roll you should consider a split early on. That's because you must aggressively fight for an anchor before the opponent has more builders in range of his board. Splitting is not going to get any safer in the future.

When the opponent has already a high anchor of his own you have even more incentive to split although it can be more dangerous and you have a constructive number.
Suppose that you get 53 early on to point behind his five anchor. If you can split you'll probably do it passing the point behind him. You can't blitz him and you can't prime him so you must get into a mutual holding game as fast as possible even if he has a stronger front position. You can't let him prime you easily.

In the first two positions your decisions are too passive for the early game and sometimes you'll face trouble because of them in the future. If you get a poor racing roll that doesn't construct, hit or maybe starts to escape the back checker, you must be dynamic with it to increase the chance of doing those things in the next rolls. There are always some risks, but if you don't play like that your moves may be riskier in the future.

In the third position your plan is to race and put some pressure on his back checker. 32 is a poor racing number. Playing 13/8 doesn't get much. Bring a builder down for more point making numbers (even though there's some duplication) and prepare to escape the back checker while preventing the builder on eight from easily making the bar or four pt. If he wants to hit you he is not starting the point he wants to make, he may not like to break the eight to merely start the three and you always have at least some return shots.
Consider this also: if you not bring a builder down and prepare to escape you're asking for the dice to cooperate with you instead of creating more good rolls for yourself while his position is not threatening.

This is what I can say. Better players may help you more.

I think, you have checked bgonline. When not, there you will see, how complicated bg opening theory has developed. It has reached the third move, not only cash game, but also AtS (At this Score). Neil Kazaross, Nack Ballard, Stick Rice and many others are posting 42k RO's to show for example a move is 0.004 better then another. I suggested, Nack Ballard should write a book like "common opening errors >0.02 and the reasons behind". He asked, if i would buy such a book. So perhaps... It is right, that every game has opening moves. But this is far to much for my blood. If i want to go to Monte Carlo every year... I have enough problems to calculate market windows. Also Falafel was curious about so much effort. Anything other fine in your game?!

Opps, there are four pointing numbers on the 20 in 697, 11, 33 as well as 31, but that does not change why I think you should split.

Thanks to your quick reactions Will, Bill, Flaccus and higonfive,

To Will:

Thanks to your reply I looked closer at the hitting numbers and I found out that there is also duplication on 35, making the 22-point and hitting on the 9 (and also unstacking to the best place with 12/20).

To Bill:

Your absolutely right that only position 1 is a problem with the question "should I split or not?" But in considering the risks and rules I see a connection between the 3 problems.

To Flaccus:

I thought your entry was the most enlightening for me. I indeed didn't consider the number of builders/attackers, for me that was the missing link.
So an extra rule could be: As long as there are 7 checkers on the other side you might consider some extra risk.

To Higonfive:

I am not so very interested in the equity numbers, but i am interested in the why of the position/best move. I try to define "rules" because these rules strangely enough make me more flexible in my approach. For instance since I learned about the term Middle -Eastern split I stopped considering this split as a good move, but digging deeper made me realize that there actually are positions in which this split is the best move. Also I saw that there are positions in which its actually good to split against extra builders when these builders are loose blots, I call this pressure-splits. So a move with a better equity for me is only the starting point for a qualitative approach.

greetings k.

In the first position, Bill Robertie's notion of efficiency plays a factor. In the early game especially, it is important to do two good things with each roll rather than one.

In your position, most of the benefits of splitting have already been achieved with the move to the 23 point. Although the major split to the 20 point may be a stronger split, the marginal gain over the minor spilt is small compared to the advantage of playing a builder down from the overstacked midpoint. Splitting alone is not as strong as splitting, unstacking and building. In Robertie's language, it is more efficient to use a roll to do two things than just one.

Incidentally, I learned that you can wrap the GNUBG ASCII board with "Code" tags (click the #), and the forum won't screw up the display.

Hi Mike,

I tried your trick,

and it works,

thanks (also for the efficiency explanation),

greetings k.

I thought I would just add a piece of tactical criteria by Phil Simborg. He writes,

"One of my favorite checker play rules of thumb is that if your opponent only has two checkers on his 8 point that should tend to make you split your back checkers. If your opponent then rolls 3/1 4/2 6/1 or 5/3, his best numbers, then at least you have a direct shot at the blot on the 8 point."

I think Flaccus was right on when he said that in the early game you should be thinking split with any non point making or hitting numbers. I believe that since the bots, we have learned to split much more often then suggested in early texts like Magriels BG.